Approximate Stream Reasoning with Incomplete State Information - - PowerPoint PPT Presentation

approximate stream reasoning with incomplete state
SMART_READER_LITE
LIVE PREVIEW

Approximate Stream Reasoning with Incomplete State Information - - PowerPoint PPT Presentation

Approximate Stream Reasoning with Incomplete State Information Fourth Stream Reasoning Workshop, Link oping, Sweden Daniel de Leng Artificial Intelligence and Integrated Computer Systems Department of Computer and Information Science Link


slide-1
SLIDE 1

Approximate Stream Reasoning with Incomplete State Information

Fourth Stream Reasoning Workshop, Link¨

  • ping, Sweden

Daniel de Leng

Artificial Intelligence and Integrated Computer Systems Department of Computer and Information Science Link¨

  • ping University, Sweden
slide-2
SLIDE 2

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Metric Temporal Logic Progression-based Runtime Verification

Introduction

1 Introduction 2 Stream Reasoning with Incomplete Information 3 Progression Graph-Based Progression 4 Summary

Daniel de Leng Link¨

  • ping University

2/19

slide-3
SLIDE 3

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Metric Temporal Logic Progression-based Runtime Verification

Introduction

Consider runtime verification for checking whether an agent is behaving in a safe manner. Example (Safety) “A robot in an unsafe state should return to a safe state within 10 seconds” Motivation: Incomplete information!

Daniel de Leng Link¨

  • ping University

3/19

slide-4
SLIDE 4

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Metric Temporal Logic Progression-based Runtime Verification

Metric Temporal Logic

We use Metric Temporal Logic (MTL) as a language for describing temporal rules that must hold. Definition (MTL syntax) The syntax for MTL is as follows for atomic propositions p ∈ Prop, temporal interval I ⊆ (0, ∞), and well-formed formulas (wffs) φ and ψ: p | ¬φ | φ ∨ ψ | φ UI ψ where I and ♦I are syntactic sugar for ‘always’ and ‘eventually’.

Daniel de Leng Link¨

  • ping University

4/19

slide-5
SLIDE 5

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Metric Temporal Logic Progression-based Runtime Verification

Progression-based Runtime Verification

Progression is an incremental syntactic rewriting procedure introduced by Bacchus and Kabanza (1996, 1998): MTL Formula + Complete State + Delay ⇒ MTL Formula φ0 = (¬p → ♦[0,10]p), s = {¬p} , ∆ = 2 φ1 = ♦[0,8]p ∧ (¬p → ♦[0,10]p)

Daniel de Leng Link¨

  • ping University

5/19

slide-6
SLIDE 6

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Incomplete States and Streams Progression Graphs

Stream Reasoning with Incomplete Information

Problem: How to perform progression with incomplete states? General idea: Apply model counting

Daniel de Leng Link¨

  • ping University

6/19

slide-7
SLIDE 7

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Incomplete States and Streams Progression Graphs

Incomplete States and Streams

Important assumptions: We keep a constant delay value (∆) and omit it from here on; An incomplete state s is a disjunctive set of complete states; A (piecewise) incomplete stream ρ is a sequence of incomplete states; We assume we have a probabilistic model of a stream denoted by a state universe Sn for every time-point n.

Daniel de Leng Link¨

  • ping University

7/19

slide-8
SLIDE 8

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Incomplete States and Streams Progression Graphs

Progression Graphs

A progression graph encodes formulas and their progressions into a graph G(χ) = (χ, V , E) such that vertices represent formulas; χ ∈ V represents the graph source formula; and labelled edges (φ, ψ, s) ∈ E iff PROGRESS(φ, s) = ψ.

⊤ ⊥ p {p} ∅ ◇[0,1] p {p} ∅ ◇[0,2] p {p} ∅ ◇[0,3] p {p} ∅ ◇[0,4] p {p} ∅ ◇[0,5] p {p} ∅

Daniel de Leng Link¨

  • ping University

8/19

slide-9
SLIDE 9

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Incomplete States and Streams Progression Graphs

Progression Graphs

Progression graphs Gn(χ) = (χ, V , E, mn) carry probability mass: m0(χ) = 1.0 (Initialization) mn(v) =

  • (v′,v,s)∈E
  • mn−1(v′) Pr[Sn = s |

sn]

  • Theorem (Soundness)

Given a progression graph Gn(χ) and a stream ρ: lim

n→∞ mn(⊤) = Pr [

ρ, t0 | = χ] , lim

n→∞ mn(⊥) = Pr [

ρ, t0 | = χ] .

Daniel de Leng Link¨

  • ping University

9/19

slide-10
SLIDE 10

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Progression Graph-Based Progression

Example (Ship Stabilisation) Suppose we have an autonomous ship with a landing deck. The ship attempts to stabilise itself according to the rule: (¬p → (♦[0,5][0,3]p)) “Whenever the ship is unstable (¬p), the ship will be stable (p) for a consecutive period of 3 minutes, within 5 minutes from having become unstable.“

Daniel de Leng Link¨

  • ping University

10/19

slide-11
SLIDE 11

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

11/19

slide-12
SLIDE 12

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

11/19

slide-13
SLIDE 13

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

11/19

slide-14
SLIDE 14

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

11/19

slide-15
SLIDE 15

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Incomplete Information

Example (Ship Stabilisation (Cont’d)) Suppose we are no longer able to measure unambiguously whether the ship is stable. Continue progression, and assume 90% stable, 10% unstable.

Daniel de Leng Link¨

  • ping University

12/19

slide-16
SLIDE 16

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-17
SLIDE 17

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-18
SLIDE 18

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-19
SLIDE 19

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-20
SLIDE 20

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-21
SLIDE 21

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-22
SLIDE 22

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-23
SLIDE 23

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression Daniel de Leng Link¨

  • ping University

13/19

slide-24
SLIDE 24

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Example: Ship Stabilisation

Example (Ship Stabilisation (Cont’d)) After 10 minutes, despite incomplete sensor readings, we know: Pr[ ρ, t0 | = (¬p → (♦[0,5][0,3]p))] ≥ 0.212680, right now based on m10(⊥), regardless of future readings.

Daniel de Leng Link¨

  • ping University

14/19

slide-25
SLIDE 25

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Approximate Progression

Approximate progression allows us to trade precision for speed and vice-versa:

1 Institute a MAX AGE for formulas; 2 Limit the size of the graph by setting a MAX NODES bound.

We may drop nodes with probability mass, thereby leaking some probability mass over time.

Daniel de Leng Link¨

  • ping University

15/19

slide-26
SLIDE 26

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Methods to reduce the graph size: MAX AGE and MAX NODES.

Daniel de Leng Link¨

  • ping University

16/19

slide-27
SLIDE 27

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Performance penalty: MAX AGE = 3

Daniel de Leng Link¨

  • ping University

17/19

slide-28
SLIDE 28

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary Complete Information Incomplete Information Approximate Progression

Precision penalty: MAX NODES = 5

Daniel de Leng Link¨

  • ping University

18/19

slide-29
SLIDE 29

Introduction Stream Reasoning with Incomplete Information Progression Graph-Based Progression Summary

Summary

Summary:

1 Classical progression assumes complete states; 2 We extended progression to handle incomplete states; 3 Progression graphs allow us to implicitly keep track of traces; 4 Approximation offers a trade-off between precision and speed.

Many interesting extensions possible!

Daniel de Leng Link¨

  • ping University

19/19